Comparing Alternative Output-Gap Estimators: A Monte Carlo Approach
The author evaluates the ability of a variety of output-gap estimators to accurately measure the output gap in a model economy. A small estimated model of the Canadian economy is used to generate artificial data. Using output and inflation data generated by this model, the author uses each output-gap estimation methodology to construct an estimate of the true output gap. He then evaluates the methodologies by comparing their respective estimates of the output gap with the true gap. The estimators are evaluated on the basis of correlations between the actual and estimated output gap, as well as the root-mean-squared estimation error. The author also varies the properties of potential output and the output gap in the data-generating process to test the robustness of his results. His findings indicate that an estimator that combines the Hodrick-Prescott filter with a Blanchard-Quah structural vector autoregression (SVAR) yields an estimate that is accurate compared with competing methods at the end-of-sample. He also finds that the performance of the SVAR relative to that of other methodologies is quite robust to violations in the identifying assumptions of the SVAR.